I decided to go all in on AI prompts for work after I watched most of my peers treat their chatbot of choice like a slightly fancier Google. There is a real edge in shipping cleaner work faster while everyone else copy-pastes the same generic prompts they saw on LinkedIn.
Over the past couple of months I built and stress-tested about 60 AI prompts for work across ChatGPT and Claude. Thirty of them earned a permanent spot in my weekly rotation. The rest got cut without much ceremony.
This post is the working library. I run a solopreneur marketing business, so most of these AI prompts for work are tuned for knowledge work, client deliverables, and small business operations. They translate to any role where you write, decide, and ship things on a deadline. Almost all of them run fine on ChatGPT, Claude, or whichever frontier model you happen to live inside.
What are the 30 AI prompts for work I use every week?
I use 30 AI prompts for work each week, grouped into six categories: communication, writing, research, decisions, strategy, and admin. The ones that survived all share a pattern. They give the model a role, the right context, a format, and an explicit way to know if the answer is good enough.
The rest of this post is the list, grouped by category, with the exact prompt language I paste in. I have spent enough time inside the GPT-5 and Claude 4 generation this year to know which patterns hold up under deadline pressure and which fall apart the second the input gets messy.
Which AI prompts for work handle communication?
Communication eats more knowledge worker time than anyone admits. I use seven of these prompts to compress that bucket from hours to minutes. Each one assumes I have already pasted in the relevant email, message, or transcript.
1. Email triage
“You are my executive assistant. Read this email thread and tell me, in three lines: what they want, what the next step is, and whether I need to reply today.”
2. Reply draft
“Write a reply in my voice (warm, direct, no jargon) that confirms [X], pushes back on [Y], and ends with a single next step. Keep it under 80 words.”
3. Slack condense
“Rewrite this message as a two-line Slack update. Lead with the decision or ask. Move detail to the second line. No greetings.”
4. Meeting prep
“I have a 30-minute meeting with a [role]. The goal is [goal]. Give me five questions to ask, in order, that move us toward a decision.”
5. Meeting recap
“Here is my raw call transcript. Pull out: decisions made, open questions, and action items with owners. Use three short lists.”
6. Difficult conversation
“I need to tell [person] [hard thing]. Give me three opening lines that are direct, kind, and assume good intent. Include what I should avoid saying.”
7. Calendar push
“Draft a polite message asking to reschedule [meeting] to next week. Keep it under 40 words. Do not over-apologize.”
The communication prompts above are the ones I trigger from muscle memory. If I have to think about how to ask, I skip them under deadline pressure. (I am still tightening the meeting prep one, far from a finished version.)
Which prompts move me from blank page to draft?
Writing is the part of the day where AI gives me the biggest leverage. These six prompts move me from blank page to shippable draft in roughly half the time. The pattern: outline first, generate from the outline, then run the same draft through three editing passes. Claude tends to win the voice match step for me, while ChatGPT often wins on headlines.
8. Outline first
“Outline a [blog post / memo / proposal] on [topic] for [audience]. Use H2s and H3s. Each H3 has a one-line summary of what it will argue.”
9. First draft from outline
“Write the first draft from the outline above. Keep paragraphs to two or three sentences. Lead each section with a direct answer.”
10. Voice match
“Rewrite the section below in my voice. My voice is: first person, short sentences, conversational, no corporate hedging. Replace any phrase a consultant would say.”
11. Tighten by 30 percent
“Cut 30 percent of the words without losing meaning. Keep the strongest line at the top of each section. Flag any sentence that lost the point.”
12. Headline test
“Give me 10 headline options for this piece. Mix curiosity, specificity, and benefit. Rank them by clarity first, click appeal second.”
13. SEO sanity check
“Here is my draft. The primary keyword is [keyword]. Tell me where it appears, where it should appear more naturally, and suggest three specific edits. Do not rewrite.”
If you only test five prompts this week, I would pick this writing batch. They compound across every kind of document I produce, from client decks to long-form essays. The tighten-by-30-percent prompt has, on its own, probably saved me more time than any other AI prompts for work in this entire list. I also covered the marketing-flavored cousins of these prompts in my ChatGPT marketing prompts library if your day-to-day is more campaign-driven than client work.
Which prompts speed up research?
Research used to be the slowest part of any project. Five prompts have cut my prep time roughly in half on most assignments. The trick is to use AI for the breadth pass, then go to primary sources for the depth pass. The current crop of models is excellent at telling you what to look at next, less reliable when you need a specific stat verbatim.
14. One-paragraph briefing
“Brief me on [topic] in one paragraph. Assume I know nothing. End with two questions an expert would still be debating.”
15. Source extraction
“From this report or article, pull the five most important data points with the page or section reference. Skip anything anecdotal.”
16. Counterargument
“I am writing about [position]. Give me the three strongest counterarguments and the best evidence behind each one. Be ruthless.”
17. Beginner glossary
“List the 10 terms a beginner needs to understand [topic]. One-line definition each. Flag any I am likely getting wrong.”
18. Compare and contrast
“Compare [option A] and [option B] across [criteria]. Use a table. Add a final row for which one wins by default and why.”
The counterargument prompt above is the one I underestimated for the longest time. AI is annoyingly good at telling me what is wrong with my own thinking when I ask it in those exact words. (I would rather find the gap in my own argument than have a client find it for me.)
Which prompts help me make decisions?
Decisions are where I used to lose the most time to second-guessing. Four prompts give me a faster way to think out loud without dragging a friend into a 40-minute voice memo. AI is not making the decision here. It is helping me see the shape of the decision faster.
19. Steel-man both sides
“Steel-man both sides of this decision: [A vs B]. Give me the best case for each. Then tell me which side has stronger evidence and why.”
20. 10-10-10
“How will this decision look in 10 minutes, 10 months, and 10 years? Be specific to my situation. No generic platitudes.”
21. Pre-mortem
“Imagine this project failed in six months. What are the three most likely reasons? What early signals should I watch for?”
22. Decision log entry
“Summarize this decision in one paragraph: context, options I considered, what I chose, and why. Format it for a decision log.”
The decision log prompt is the one I almost cut and then kept. Future-me reading a one-paragraph summary of why I made a call is worth more than the 30 seconds it costs to ask.
Which prompts handle strategy work?
Strategy work usually starts with a fuzzy idea and ends with something I can show a client or a team. These four prompts do most of the heavy lifting between “vague hunch” and “one-pager I can send”. The framing prompt below is the one I keep coming back to whenever I get stuck on a problem.
23. Frame the problem three ways
“Restate this problem three different ways: as a customer problem, as an operations problem, and as a positioning problem. Which framing unlocks the most useful next move?”
24. Quick SWOT
“Run a quick SWOT on [project or business]. Keep it to three bullets per quadrant. Then tell me which threat I am underestimating.”
25. Offer stress test
“Here is my offer or pitch: [paste]. Tell me what is unclear, what is unbelievable, and what is missing. Be honest, not nice.”
26. Quarterly plan in a page
“Help me draft a one-page plan for next quarter. Three priorities. One success metric each. One thing I will deliberately stop doing.”
If you run your own business or operate a small team, the strategy prompts above overlap heavily with the broader templates I keep in my AI prompts for business library. The framing prompt is the one I reach for most, especially when I have been staring at the same problem for an hour and need a clean angle on it.
Which prompts handle admin and operations?
The admin layer is where AI quietly saves me hours every week. Four prompts handle the parts of the job no one wants to do. None of these are glamorous. All of them remove friction.
27. SOP from messy notes
“Turn these messy notes into a step-by-step SOP. Number the steps. Add a quick check at the end of each step so someone could follow it without me in the room.”
28. Recurring task checklist
“Build me a checklist for [recurring task]. Keep it under 10 items. Flag the steps where I usually slip.”
29. Inbox sort
“Group my unread emails by sender intent: action needed, FYI, newsletter, spam. Suggest a one-click action for each group.”
30. Data cleanup pass
“Clean this messy data. Standardize columns, format dates, and flag any rows that look wrong. Do not delete anything. Just mark them.”
The SOP prompt is the sneaky favorite. I have used it to document everything from invoice cycles to monthly client reporting. Each pass turns into a Loom-and-checklist combo that I can hand off later if I ever hire help.
What makes an AI prompt for work actually useful?
A useful AI prompt for work has four things: a role, the right context, a clear format, and a success test. Drop any one of those and the output drifts toward generic. This is true on every model I have tested, from the small fast ones to the frontier flagships.
I noticed something else after testing this many prompts. The ones that survived my weekly rotation were rarely the cleverest ones. They were the ones I could trigger from muscle memory. The bar for “great prompt” is lower than people think, and “the prompt I actually use” is the only one that compounds.
The shortest path to a useful prompt is rewriting one you stole from someone else until it sounds like the way you would actually ask a smart coworker for help. If your prompt reads like a software contract, it is too long. If it reads like a Slack DM, it is closer to right.
One more thing. The models have gotten faster and more reliable across both the GPT-5 and Claude 4 generations this year, but the prompts that worked in 2024 still work today. The big shift has been less in the models and more in how I batch my work around the prompts. (My AI productivity tools stack goes deeper on that side of the workflow if you want the broader picture.)
The Bigger Pattern Emerging in How I Use AI Prompts for Work
The pattern I keep noticing is that AI prompts for work do not save me time by writing faster. They save me time by removing the staring-at-blank-screen step. Once I have something to react to, I move quickly. That holds whether I am inside ChatGPT, Claude, or anything else.
That changes what I value in a prompt. I stopped chasing one-shot magic and started building small, repeatable prompts that get me to a usable draft fast. The polish stays mine. The blank page is gone.
If you are building your own library of AI prompts for work, start with the five tasks you do every week and write a prompt for each. Test each one for a week. Cut the ones you do not reach for. Keep the ones you trigger without thinking.
And honestly, despite how generic most AI prompt lists feel, a small library you actually use beats a polished one you never open. Mine is still a moving target (far from a finished system) and that is fine. The prompts are tools, and the tools get sharper the more I use them.